Upload 6 files

chatAgentPdf/utils/ask_human.py

@@ -0,0 +1,32 @@
+""" 
+Custom Langchain tool to ask human
+"""
+
+import time
+
+import streamlit as st
+from langchain.tools.base import BaseTool
+
+
+class CustomAskHumanTool(BaseTool):
+    """Tool that asks user for input."""
+
+    name = "AskHuman"
+    description = """Use this tool if you don't find a specific answer using KendraRetrievalTool.\
+Ask the human to clarify the question or provide the missing information.\
+The input should be a question for the human."""
+
+    def _run(
+        self,
+        query: str,
+        run_manager=None,
+    ) -> str:
+        if "user_answer" not in st.session_state:
+            answer_container = st.chat_message("assistant", avatar="🦜")
+            answer_container.write(query)
+
+            answer = st.text_input("Enter your answer", key="user_answer")
+            while answer == "":
+                time.sleep(1)
+
+        return st.session_state["user_answer"]

chatAgentPdf/utils/model_params.py

@@ -0,0 +1,51 @@
+"""
+Utilities for modeling
+"""
+
+
+def get_model_params(
+    model_id: str,
+    params: dict,
+) -> dict:
+    """
+    Set up a dictionary with model parameters named appropriately for Bedrock
+
+    Parameters
+    ----------
+    model_id : str
+        Model name
+    params : dict
+        Inference parameters
+
+    Returns
+    -------
+    dict
+        _description_
+    """
+
+    model_params = {}
+
+    # name parameters based on the model id
+    if model_id.startswith("amazon"):
+        model_params = {
+            "maxTokenCount": params["answer_length"],
+            "stopSequences": params["stop_words"],
+            "temperature": params["temperature"],
+            "topP": params["top_p"],
+        }
+    elif model_id.startswith("anthropic"):
+        model_params = {
+            "max_tokens_to_sample": params["answer_length"],
+            "stop_sequences": params["stop_words"],
+            "temperature": params["temperature"],
+            "top_p": params["top_p"],
+        }
+    elif model_id.startswith("ai21"):
+        model_params = {
+            "maxTokens": params["answer_length"],
+            "stopSequences": params["stop_words"],
+            "temperature": params["temperature"],
+            "topP": params["top_p"],
+        }
+
+    return model_params

chatAgentPdf/utils/prompts.py

@@ -0,0 +1,49 @@
+""" 
+Custom Langchain prompt templates
+"""
+
+from langchain.prompts import PromptTemplate
+
+
+def create_qa_prompt() -> PromptTemplate:
+    """
+    Prompt for retrieval QA chain
+    """
+
+    template = """\n\nHuman: Use the following pieces of context to answer the question at the end.
+
+{context}
+
+Question: {question}
+
+\n\nAssistant:
+Answer:"""
+
+    return PromptTemplate(template=template, input_variables=["context", "question"])
+
+
+def create_agent_prompt() -> PromptTemplate:
+    """
+    Prompt for the agent
+    """
+
+    prefix = """\n\nHuman: Answer the following questions as best you can. You have access to the following tools:"""
+
+    format_instructions = """Use the following format:
+
+Question: the input question you must answer
+Thought: you should always think about what to do
+Action: the action to take, should be one of [{tool_names}]
+Action Input: the input to the action
+Observation: the result of the action
+... (this Thought/Action/Action Input/Observation can repeat N times)
+Thought: I now know the final answer
+Final Answer: the final answer to the original input question"""
+
+    suffix = """Begin!
+Question: {input}
+\n\nAssistant:
+Thought: {agent_scratchpad}
+"""
+
+    return prefix, format_instructions, suffix